Convolutional Neural Networks

Project: Write an Algorithm for a Dog Identification App


In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the Jupyter Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this Jupyter notebook.


Why We're Here

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Write your Algorithm
  • Step 6: Test Your Algorithm

Step 0: Import Datasets

Make sure that you've downloaded the required human and dog datasets:

Note: if you are using the Udacity workspace, you DO NOT need to re-download these - they can be found in the /data folder as noted in the cell below.

  • Download the dog dataset. Unzip the folder and place it in this project's home directory, at the location /dog_images.

  • Download the human dataset. Unzip the folder and place it in the home directory, at location /lfw.

Note: If you are using a Windows machine, you are encouraged to use 7zip to extract the folder.

In the code cell below, we save the file paths for both the human (LFW) dataset and dog dataset in the numpy arrays human_files and dog_files.

In [116]:
import numpy as np
from glob import glob

# load filenames for human and dog images
human_files = np.array(glob("/data/lfw/*/*"))
dog_files = np.array(glob("/data/dog_images/*/*/*"))

# print number of images in each dataset
print('There are %d total human images.' % len(human_files))
print('There are %d total dog images.' % len(dog_files))
There are 13233 total human images.
There are 8351 total dog images.

Step 1: Detect Humans

In this section, we use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images.

OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory. In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [117]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')

# load color (BGR) image
img = cv2.imread(human_files[0])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Number of faces detected: 1

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [118]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer: (You can print out your results and/or write your percentages in this cell)

In [119]:
from tqdm import tqdm

human_files_short = human_files[:100]
dog_files_short = dog_files[:100]

#-#-# Do NOT modify the code above this line. #-#-#

## TODO: Test the performance of the face_detector algorithm 
## on the images in human_files_short and dog_files_short.
detected_in_human = 0
detected_in_dogs = 0
for i in range(100):
    detected_in_human += int(face_detector(human_files_short[i]))
    detected_in_dogs += int(face_detector(dog_files_short[i]))
print("Performance on first 100 human images (detected faces): " + str((detected_in_human/100) * 100) + "%")
print("Performance on first 100 dog images (detected faces): " + str((detected_in_dogs/100) * 100) + "%")
Performance on first 100 human images (detected faces): 98.0%
Performance on first 100 dog images (detected faces): 17.0%

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.

In [120]:
### (Optional) 
### TODO: Test performance of anotherface detection algorithm.
### Feel free to use as many code cells as needed.

Step 2: Detect Dogs

In this section, we use a pre-trained model to detect dogs in images.

Obtain Pre-trained VGG-16 Model

The code cell below downloads the VGG-16 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories.

In [121]:
import torch
import torchvision.models as models

# define VGG16 model
VGG16 = models.vgg16(pretrained=True)

# check if CUDA is available
use_cuda = torch.cuda.is_available()

# move model to GPU if CUDA is available
if use_cuda:
    VGG16 = VGG16.cuda()

Given an image, this pre-trained VGG-16 model returns a prediction (derived from the 1000 possible categories in ImageNet) for the object that is contained in the image.

(IMPLEMENTATION) Making Predictions with a Pre-trained Model

In the next code cell, you will write a function that accepts a path to an image (such as 'dogImages/train/001.Affenpinscher/Affenpinscher_00001.jpg') as input and returns the index corresponding to the ImageNet class that is predicted by the pre-trained VGG-16 model. The output should always be an integer between 0 and 999, inclusive.

Before writing the function, make sure that you take the time to learn how to appropriately pre-process tensors for pre-trained models in the PyTorch documentation.

In [122]:
from PIL import Image, ImageFile
import torchvision.transforms as transforms

# Allow truncated images in dataset, otherwise PIL will raise an error
ImageFile.LOAD_TRUNCATED_IMAGES = True

def VGG16_predict(img_path):
    '''
    Use pre-trained VGG-16 model to obtain index corresponding to 
    predicted ImageNet class for image at specified path
    
    Args:
        img_path: path to an image
        
    Returns:
        Index corresponding to VGG-16 model's prediction
    '''
    
    ## TODO: Complete the function.
    ## Load and pre-process an image from the given img_path
    ## Return the *index* of the predicted class for that image
    img = Image.open(img_path)
    # Resize the input image to the network's input size
    # Normalize the image and turn it to Tensor
    transform1 = transforms.Compose([transforms.Resize(size=(224,224)),
                        transforms.ToTensor(),
                        transforms.Normalize((0.485, 0.456, 0.406), 
                                             (0.229, 0.224, 0.225))])
    # discard the transparent, alpha channel (that's the :3)
    img = transform1(img)[:3,:,:].unsqueeze(0)
    if use_cuda:
        img = img.cuda()
    # get model percentage outputs
    prediction = VGG16.forward(img)
    # get highest prediction (index)
    prediction = np.argmax(prediction.cpu().detach().numpy(), axis=-1)
    return prediction # predicted class index

(IMPLEMENTATION) Write a Dog Detector

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained VGG-16 model, we need only check if the pre-trained model predicts an index between 151 and 268 (inclusive).

Use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [123]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    ## TODO: Complete the function.
    index = VGG16_predict(img_path)
    return (index[0] >= 151 and index[0] <= 268) # true/false

(IMPLEMENTATION) Assess the Dog Detector

Question 2: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer:

In [124]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
d_detected_in_dogs = 0
d_detected_in_humans = 0
for dog_im in dog_files_short:
    d_detected_in_dogs += int(dog_detector(dog_im))

for human_im in human_files_short:
    d_detected_in_humans += int(dog_detector(human_im))

print("Performance on first human images (detected dogs): " + str((d_detected_in_humans/len(human_files_short)) * 100) + "%")
print("Performance on first 100 dog images (detected dogs): " + str((d_detected_in_dogs/len(dog_files_short)) * 100) + "%")
Performance on first human images (detected dogs): 0.0%
Performance on first 100 dog images (detected dogs): 100.0%

We suggest VGG-16 as a potential network to detect dog images in your algorithm, but you are free to explore other pre-trained networks (such as Inception-v3, ResNet-50, etc). Please use the code cell below to test other pre-trained PyTorch models. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.

In [125]:
### (Optional) 
### TODO: Report the performance of another pre-trained network.
### Feel free to use as many code cells as needed.

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 10%. In Step 4 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have trouble distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador Black Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset

Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dog_images/train, dog_images/valid, and dog_images/test, respectively). You may find this documentation on custom datasets to be a useful resource. If you are interested in augmenting your training and/or validation data, check out the wide variety of transforms!

In [11]:
import os
from torchvision import datasets
import torchvision.transforms as transforms
from torch.utils.data.sampler import SubsetRandomSampler
### TODO: Write data loaders for training, validation, and test sets
## Specify appropriate transforms, and batch_sizes

# number of subprocesses to use for data loading
num_workers = 0
# how many samples per batch to load
batch_size = 64

# convert data to a normalized torch.FloatTensor
train_transform = transforms.Compose([transforms.Resize(size=(224, 224)),
                                transforms.RandomHorizontalFlip(0.2),
                                transforms.RandomVerticalFlip(0.2),
                                transforms.RandomRotation(30),
                                transforms.ToTensor(),
                                transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])

test_transform = transforms.Compose([transforms.Resize(size=(224, 224)),
                                    transforms.ToTensor(),
                                    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])


train_dir = "/data/dog_images/train"
test_dir = "/data/dog_images/test"
valid_dir = "/data/dog_images/valid"

train_data = datasets.ImageFolder(train_dir, transform=train_transform)
test_data = datasets.ImageFolder(test_dir, transform=test_transform)
valid_data = datasets.ImageFolder(valid_dir, transform=test_transform)

# prepare data loaders
train_loader = torch.utils.data.DataLoader(train_data, batch_size=batch_size, 
                                           num_workers=num_workers, shuffle=True)
test_loader = torch.utils.data.DataLoader(test_data, batch_size=batch_size, 
                                          num_workers=num_workers, shuffle=False)
valid_loader = torch.utils.data.DataLoader(valid_data, batch_size=batch_size, 
                                          num_workers=num_workers, shuffle=False)

loaders_scratch = {'train':train_loader,
                   'valid':valid_loader,
                   'test':valid_loader}

Question 3: Describe your chosen procedure for preprocessing the data.

  • How does your code resize the images (by cropping, stretching, etc)? What size did you pick for the input tensor, and why?
  • Did you decide to augment the dataset? If so, how (through translations, flips, rotations, etc)? If not, why not?

Answer:

  • All images are resized using transforms.Resize which stretches the images or squeshes it to shape of 224x224 as this is the size my network input will have.
  • The data is normalized using the transforms.Normalize method and converted to torch Tensor.
  • The train dataset is augmented using random horziontal flipping, vertical flipping with probability 0.2. also random rotations up to 30 degrees.

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. Use the template in the code cell below.

In [12]:
import torch.nn as nn
import torch.nn.functional as F

# define the CNN architecture
class Net(nn.Module):
    ### TODO: choose an architecture, and complete the class
    def __init__(self):
        super(Net, self).__init__()
        ## Define layers of a CNN
        # Sees 224x224x3 input
        self.conv1 = nn.Conv2d(3, 32, 3, padding=1)
        self.conv1_bn = nn.BatchNorm2d(32)
        # Sees 112x112x32 input
        self.conv2 = nn.Conv2d(32, 64, 3, padding=1)
        self.conv2_bn = nn.BatchNorm2d(64)
        # Sees 56x56x64 input
        self.conv3 = nn.Conv2d(64, 128, 3, padding=1)
        self.conv3_bn = nn.BatchNorm2d(128)
        # Sees 28x28x128 input
        self.conv4 = nn.Conv2d(128, 256, 3, padding=1)
        self.conv4_bn = nn.BatchNorm2d(256)
        # Sees 14x14x256 input
        self.conv5 = nn.Conv2d(256, 512, 3, padding=1)
        self.conv5_bn = nn.BatchNorm2d(512)
        # max pooling layer
        self.pool = nn.MaxPool2d(2, 2)
        
        # Sees 7x7x512 input
        self.fc1 = nn.Linear(7 * 7 * 512, 512)
        self.fc2 = nn.Linear(512, 512)
        self.fc3 = nn.Linear(512, 133)
        
        self.dropout = nn.Dropout(0.2)
        
        
    def forward(self, x):
        ## Define forward behavior
        # add sequence of convolutional and max pooling layers
        x = self.pool(F.relu(self.conv1_bn(self.conv1(x))))
        x = self.pool(F.relu(self.conv2_bn(self.conv2(x))))
        x = self.pool(F.relu(self.conv3_bn(self.conv3(x))))
        x = self.pool(F.relu(self.conv4_bn(self.conv4(x))))
        x = self.pool(F.relu(self.conv5_bn(self.conv5(x))))
        # flatten image input
        x = x.view(-1, 7 * 7 * 512)
        x = self.dropout(x)
        # add 1st hidden layer, with relu activation function
        x = F.relu(self.fc1(x))
        # add 2nd hidden layer, with relu activation function
        x = F.relu(self.fc2(x))
        # output unit with no activation function
        x = self.fc3(x)

        
        return x

#-#-# You so NOT have to modify the code below this line. #-#-#

# instantiate the CNN
model_scratch = Net()

print(model_scratch)

# move tensors to GPU if CUDA is available
if use_cuda:
    model_scratch.cuda()
Net(
  (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv1_bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv2_bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv3): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv3_bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv4): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv4_bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv5): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv5_bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (fc1): Linear(in_features=25088, out_features=512, bias=True)
  (fc2): Linear(in_features=512, out_features=512, bias=True)
  (fc3): Linear(in_features=512, out_features=133, bias=True)
  (dropout): Dropout(p=0.2)
)

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step.

Answer: I chose my network based on trial and error, I tried different combinations of the number of feature maps in each layer and the number of fully connected layers. Failed trials:

  • 4 Convolution layers (64 feature maps then 32 then 16 then 8) followed by 2 fully connected layers. (724x392 --> 392x133)
  • Same like the previous but with less feature maps and less nodes in the fully connected layers
  • I tried testing the above without dropout, still the network wouldn't improve or train at all.

I then done my research and found out that most classification CNN problems start with small amount of feature maps then ends with very large number of feature maps. So I tried 5 Convolution layers (32 feature maps then 64 then 128 then 256 then 512) followed by 3 fully connected layers. (25088x512 --> 512x512 --> 512x133), the model started to train and improve but still very very slowly. I then tried to normalize the input images which led to the model improving much more than before!

Explanation of final design: conv1(3, 32) --> maxpool --> conv2(32, 64) --> maxpool --> conv3(64, 128) --> maxpool --> conv4(128, 256) --> maxpool --> conv5(256, 512) --> maxpool --> flatten(input_size=25088) --> hidden_layer1 (512 nodes) --> hidden_layer2(512 nodes) --> output(133 classes)

  • I used 5 convolution layers with a maxpooling layer after each convolution layer to decrease the dimensions of the propagating data while increasing the feature maps (deeper level --> more feature maps)
  • I used ReLU activation function after each layer
  • My convolution layers had a padding of (1,1) ensuring the output dimensions (2D dimensions) stayed the same
  • I also used batch normalization after each layer to increase training accuracy and to make the model converge faster
  • I used dropout with the fully connected layers to ensure no overfitting happens
  • I tested out different combinations of the fully connected layers number of nodes until my model improves

(IMPLEMENTATION) Specify Loss Function and Optimizer

Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_scratch, and the optimizer as optimizer_scratch below.

In [13]:
import torch.optim as optim
from workspace_utils import active_session

### TODO: select loss function
criterion_scratch = nn.CrossEntropyLoss()

### TODO: select optimizer
optimizer_scratch = optim.Adam(model_scratch.parameters(), lr=0.003)

(IMPLEMENTATION) Train and Validate the Model

Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_scratch.pt'.

In [14]:
def train(n_epochs, loaders, model, optimizer, criterion, use_cuda, save_path):
    """returns trained model"""
    # initialize tracker for minimum validation loss
    valid_loss_min = np.Inf 
    for epoch in range(1, n_epochs+1):
        # initialize variables to monitor training and validation loss
        train_loss = 0.0
        valid_loss = 0.0
        
        ###################
        # train the model #
        ###################
        model.train()
        for batch_idx, (data, target) in enumerate(loaders['train']):
            # move to GPU
            if use_cuda:
                data, target = data.cuda(), target.cuda()
            ## find the loss and update the model parameters accordingly
            ## record the average training loss, using something like
            ## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
            
            # clear the gradients of all optimized variables
            optimizer.zero_grad()
            
            # forward pass: compute predicted outputs by passing inputs to the model
            output = model(data)
            
            # calculate loss
            loss = criterion(output, target)
            
            # backpropagate: compute gradient of the loss with respect to model parameters
            loss.backward()
            
            # perform a single optimization step (parameter update)
            optimizer.step()
            
            # update training loss
            train_loss += ((1 / (batch_idx + 1)) * (loss.data - train_loss))
        ######################    
        # validate the model #
        ######################
        model.eval()
        for batch_idx, (data, target) in enumerate(loaders['valid']):
            with torch.no_grad():
            # move to GPU
                if use_cuda:
                    data, target = data.cuda(), target.cuda()
            ## update the average validation loss
           
                output = model(data)
                loss = criterion(output, target)
                valid_loss += ((1 / (batch_idx + 1)) * (loss.data - valid_loss))

            
        # print training/validation statistics 
        print('Epoch: {} \tTraining Loss: {:.6f} \tValidation Loss: {:.6f}'.format(
            epoch, 
            train_loss,
            valid_loss
            ))
        
        ## TODO: save the model if validation loss has decreased
        if(valid_loss <= valid_loss_min):
            print('Validation loss decreased ({:.6f} --> {:.6f}).  Saving model ...'.format(valid_loss_min, valid_loss))
            torch.save(model.state_dict(), save_path)
            valid_loss_min = valid_loss    
    # return trained model
    return model


# train the model
with active_session():
    model_scratch = train(30, loaders_scratch, model_scratch, optimizer_scratch, 
                      criterion_scratch, use_cuda, 'model_scratch.pt')

# load the model that got the best validation accuracy
model_scratch.load_state_dict(torch.load('model_scratch.pt'))
Epoch: 1 	Training Loss: 5.143497 	Validation Loss: 4.747413
Validation loss decreased (inf --> 4.747413).  Saving model ...
Epoch: 2 	Training Loss: 4.674192 	Validation Loss: 4.733571
Validation loss decreased (4.747413 --> 4.733571).  Saving model ...
Epoch: 3 	Training Loss: 4.477253 	Validation Loss: 4.385110
Validation loss decreased (4.733571 --> 4.385110).  Saving model ...
Epoch: 4 	Training Loss: 4.329006 	Validation Loss: 4.466588
Epoch: 5 	Training Loss: 4.255403 	Validation Loss: 4.377531
Validation loss decreased (4.385110 --> 4.377531).  Saving model ...
Epoch: 6 	Training Loss: 4.197235 	Validation Loss: 4.240879
Validation loss decreased (4.377531 --> 4.240879).  Saving model ...
Epoch: 7 	Training Loss: 4.114139 	Validation Loss: 4.190572
Validation loss decreased (4.240879 --> 4.190572).  Saving model ...
Epoch: 8 	Training Loss: 4.059108 	Validation Loss: 4.838166
Epoch: 9 	Training Loss: 3.973217 	Validation Loss: 4.036821
Validation loss decreased (4.190572 --> 4.036821).  Saving model ...
Epoch: 10 	Training Loss: 3.917277 	Validation Loss: 4.092797
Epoch: 11 	Training Loss: 3.834216 	Validation Loss: 4.120269
Epoch: 12 	Training Loss: 3.760523 	Validation Loss: 4.160929
Epoch: 13 	Training Loss: 3.676308 	Validation Loss: 4.166405
Epoch: 14 	Training Loss: 3.611569 	Validation Loss: 3.914059
Validation loss decreased (4.036821 --> 3.914059).  Saving model ...
Epoch: 15 	Training Loss: 3.544873 	Validation Loss: 3.923591
Epoch: 16 	Training Loss: 3.487895 	Validation Loss: 3.700078
Validation loss decreased (3.914059 --> 3.700078).  Saving model ...
Epoch: 17 	Training Loss: 3.417266 	Validation Loss: 3.686999
Validation loss decreased (3.700078 --> 3.686999).  Saving model ...
Epoch: 18 	Training Loss: 3.352180 	Validation Loss: 3.702005
Epoch: 19 	Training Loss: 3.279117 	Validation Loss: 3.549574
Validation loss decreased (3.686999 --> 3.549574).  Saving model ...
Epoch: 20 	Training Loss: 3.195750 	Validation Loss: 4.721534
Epoch: 21 	Training Loss: 3.158146 	Validation Loss: 3.448131
Validation loss decreased (3.549574 --> 3.448131).  Saving model ...
Epoch: 22 	Training Loss: 3.077981 	Validation Loss: 3.446328
Validation loss decreased (3.448131 --> 3.446328).  Saving model ...
Epoch: 23 	Training Loss: 3.036789 	Validation Loss: 3.321361
Validation loss decreased (3.446328 --> 3.321361).  Saving model ...
Epoch: 24 	Training Loss: 2.941941 	Validation Loss: 3.507207
Epoch: 25 	Training Loss: 2.889648 	Validation Loss: 3.287521
Validation loss decreased (3.321361 --> 3.287521).  Saving model ...
Epoch: 26 	Training Loss: 2.838379 	Validation Loss: 3.319073
Epoch: 27 	Training Loss: 2.812760 	Validation Loss: 3.366081
Epoch: 28 	Training Loss: 2.711158 	Validation Loss: 3.493451
Epoch: 29 	Training Loss: 2.672404 	Validation Loss: 3.447702
Epoch: 30 	Training Loss: 2.599290 	Validation Loss: 3.271215
Validation loss decreased (3.287521 --> 3.271215).  Saving model ...

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 10%.

In [15]:
def test(loaders, model, criterion, use_cuda):

    # monitor test loss and accuracy
    test_loss = 0.
    correct = 0.
    total = 0.

    model.eval()
    for batch_idx, (data, target) in enumerate(loaders['test']):
        # move to GPU
        if use_cuda:
            data, target = data.cuda(), target.cuda()
        # forward pass: compute predicted outputs by passing inputs to the model
        output = model(data)
        # calculate the loss
        loss = criterion(output, target)
        # update average test loss 
        test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))
        # convert output probabilities to predicted class
        pred = output.data.max(1, keepdim=True)[1]
        # compare predictions to true label
        correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())
        total += data.size(0)
            
    print('Test Loss: {:.6f}\n'.format(test_loss))

    print('\nTest Accuracy: %2d%% (%2d/%2d)' % (
        100. * correct / total, correct, total))

# call test function    
with active_session():
    test(loaders_scratch, model_scratch, criterion_scratch, use_cuda)
Test Loss: 3.271215


Test Accuracy: 19% (159/835)

Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset

Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dogImages/train, dogImages/valid, and dogImages/test, respectively).

If you like, you are welcome to use the same data loaders from the previous step, when you created a CNN from scratch.

In [23]:
## TODO: Specify data loaders
import torch
import torchvision.models as models
import os
from torchvision import datasets
import torchvision.transforms as transforms
from torch.utils.data.sampler import SubsetRandomSampler
from PIL import Image, ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
### TODO: Write data loaders for training, validation, and test sets
## Specify appropriate transforms, and batch_sizes

# number of subprocesses to use for data loading
num_workers = 0
# how many samples per batch to load
batch_size = 64

# convert data to a normalized torch.FloatTensor
train_transform = transforms.Compose([transforms.Resize(size=(224, 224)),
                                transforms.RandomHorizontalFlip(0.2),
                                transforms.RandomVerticalFlip(0.2),
                                transforms.RandomRotation(30),
                                transforms.ToTensor(),
                                transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])

test_transform = transforms.Compose([transforms.Resize(size=(224, 224)),
                                    transforms.ToTensor(),
                                    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])


train_dir = "/data/dog_images/train"
test_dir = "/data/dog_images/test"
valid_dir = "/data/dog_images/valid"

train_data = datasets.ImageFolder(train_dir, transform=train_transform)
test_data = datasets.ImageFolder(test_dir, transform=test_transform)
valid_data = datasets.ImageFolder(valid_dir, transform=test_transform)

# prepare data loaders
train_loader = torch.utils.data.DataLoader(train_data, batch_size=batch_size, 
                                           num_workers=num_workers, shuffle=True)
test_loader = torch.utils.data.DataLoader(test_data, batch_size=batch_size, 
                                          num_workers=num_workers, shuffle=False)
valid_loader = torch.utils.data.DataLoader(valid_data, batch_size=batch_size, 
                                          num_workers=num_workers, shuffle=False)

loaders_transfer = {'train':train_loader,
                   'valid':valid_loader,
                   'test':valid_loader}

(IMPLEMENTATION) Model Architecture

Use transfer learning to create a CNN to classify dog breed. Use the code cell below, and save your initialized model as the variable model_transfer.

In [24]:
import torchvision.models as models
import torch.nn as nn

## TODO: Specify model architecture 
model_transfer = models.resnet50(pretrained=True)
for param in model_transfer.parameters():
    param.requires_grad = False
In [25]:
# define my new fully connected part of the network
fc = nn.Sequential(nn.Linear(2048, 512),
                   nn.BatchNorm1d(512),
                   nn.ReLU(),
                   nn.Dropout(0.2),
                   nn.Linear(512, 133))
# set the new classifier
model_transfer.fc = fc


# check if CUDA is available
use_cuda = torch.cuda.is_available()

if use_cuda:
    model_transfer = model_transfer.cuda()

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer:

  • First I tested the VGG16 performance along with different variatons of the number of nodes/layers of the fully connected layer, at best it reached 63% accuracy in testing so I decided to try out a more complicated pretrained network which is ResNet50.
  • I tried different variations of the fully connected layer, I tried 3 hidden layers which got around 68% accuracy, tried no hidden layers but just flattening the output from ResNet50 and then the output layer and the model got around 70% accuracy.
  • I then tried one hidden layer of 1024 nodes which led to around 69% accuracy.
  • I then settled on one hidden layer of 512 nodes which led to 71% accuracy.
  • I added dropout to avoid overfitting the fully connected layer since the dataset is small.
  • I used an Adam optimizer, I was tempted to test out the SGD optimizer but decided to stick to the Adam optimizer.
  • I tried adding batch normalization for the fully connected layers and the performance jumped 10% !!, to 81% accuracy.

(IMPLEMENTATION) Specify Loss Function and Optimizer

Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_transfer, and the optimizer as optimizer_transfer below.

In [26]:
import torch.optim as optim
criterion_transfer = nn.CrossEntropyLoss()
optimizer_transfer = optim.Adam(model_transfer.fc.parameters(), lr=0.001)

(IMPLEMENTATION) Train and Validate the Model

Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_transfer.pt'.

In [27]:
def train(n_epochs, loaders, model, optimizer, criterion, use_cuda, save_path):
    """returns trained model"""
    # initialize tracker for minimum validation loss
    valid_loss_min = np.Inf 
    
    for epoch in range(1, n_epochs+1):
        # initialize variables to monitor training and validation loss
        train_loss = 0.0
        valid_loss = 0.0
        
        ###################
        # train the model #
        ###################
        model.train()
        for batch_idx, (data, target) in enumerate(loaders['train']):
            # move to GPU
            if use_cuda:
                data, target = data.cuda(), target.cuda()
            ## find the loss and update the model parameters accordingly
            ## record the average training loss, using something like
            ## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
            
            # clear the gradients of all optimized variables
            optimizer.zero_grad()
            
            # forward pass: compute predicted outputs by passing inputs to the model
            output = model(data)
            
            # calculate loss
            loss = criterion(output, target)
            
            # backpropagate: compute gradient of the loss with respect to model parameters
            loss.backward()
            
            # perform a single optimization step (parameter update)
            optimizer.step()
            
            # update training loss
            train_loss += ((1 / (batch_idx + 1)) * (loss.data - train_loss))
        ######################    
        # validate the model #
        ######################
        model.eval()
        for batch_idx, (data, target) in enumerate(loaders['valid']):
            with torch.no_grad():
            # move to GPU
                if use_cuda:
                    data, target = data.cuda(), target.cuda()
            ## update the average validation loss
           
                output = model(data)
                loss = criterion(output, target)
                valid_loss += ((1 / (batch_idx + 1)) * (loss.data - valid_loss))

            
        # print training/validation statistics 
        print('Epoch: {} \tTraining Loss: {:.6f} \tValidation Loss: {:.6f}'.format(
            epoch, 
            train_loss,
            valid_loss
            ))
        
        ## TODO: save the model if validation loss has decreased
        if(valid_loss <= valid_loss_min):
            print('Validation loss decreased ({:.6f} --> {:.6f}).  Saving model ...'.format(valid_loss_min, valid_loss))
            torch.save(model.state_dict(), save_path)
            valid_loss_min = valid_loss
            
    # return trained model
    return model
In [28]:
def test(loaders, model, criterion, use_cuda):

    # monitor test loss and accuracy
    test_loss = 0.
    correct = 0.
    total = 0.

    model.eval()
    for batch_idx, (data, target) in enumerate(loaders['test']):
        # move to GPU
        if use_cuda:
            data, target = data.cuda(), target.cuda()
        # forward pass: compute predicted outputs by passing inputs to the model
        output = model(data)
        # calculate the loss
        loss = criterion(output, target)
        # update average test loss 
        test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))
        # convert output probabilities to predicted class
        pred = output.data.max(1, keepdim=True)[1]
        # compare predictions to true label
        correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())
        total += data.size(0)
            
    print('Test Loss: {:.6f}\n'.format(test_loss))

    print('\nTest Accuracy: %2d%% (%2d/%2d)' % (
        100. * correct / total, correct, total))
In [29]:
# train the model
model_transfer = train(10, loaders_transfer, model_transfer, optimizer_transfer, criterion_transfer, use_cuda, 'model_transfer.pt')# train(n_epochs, loaders_transfer, model_transfer, optimizer_transfer, criterion_transfer, use_cuda, 'model_transfer.pt')
Epoch: 1 	Training Loss: 2.791712 	Validation Loss: 1.260443
Validation loss decreased (inf --> 1.260443).  Saving model ...
Epoch: 2 	Training Loss: 1.288096 	Validation Loss: 0.825384
Validation loss decreased (1.260443 --> 0.825384).  Saving model ...
Epoch: 3 	Training Loss: 0.978946 	Validation Loss: 0.663412
Validation loss decreased (0.825384 --> 0.663412).  Saving model ...
Epoch: 4 	Training Loss: 0.864386 	Validation Loss: 0.634121
Validation loss decreased (0.663412 --> 0.634121).  Saving model ...
Epoch: 5 	Training Loss: 0.779479 	Validation Loss: 0.610428
Validation loss decreased (0.634121 --> 0.610428).  Saving model ...
Epoch: 6 	Training Loss: 0.691288 	Validation Loss: 0.589861
Validation loss decreased (0.610428 --> 0.589861).  Saving model ...
Epoch: 7 	Training Loss: 0.650686 	Validation Loss: 0.565237
Validation loss decreased (0.589861 --> 0.565237).  Saving model ...
Epoch: 8 	Training Loss: 0.626086 	Validation Loss: 0.577060
Epoch: 9 	Training Loss: 0.596471 	Validation Loss: 0.574488
Epoch: 10 	Training Loss: 0.551083 	Validation Loss: 0.537685
Validation loss decreased (0.565237 --> 0.537685).  Saving model ...
In [126]:
# load the model that got the best validation accuracy (uncomment the line below)
model_transfer.load_state_dict(torch.load('model_transfer.pt'))

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 60%.

In [31]:
test(loaders_transfer, model_transfer, criterion_transfer, use_cuda)
Test Loss: 0.537685


Test Accuracy: 82% (685/835)

(IMPLEMENTATION) Predict Dog Breed with the Model

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan hound, etc) that is predicted by your model.

In [127]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.

# list of class names by index, i.e. a name can be accessed like class_names[0]
class_names = [item[4:].replace("_", " ") for item in train_data.classes]

def predict_breed_transfer(img_path):
    # load the image and return the predicted breed
    img = Image.open(img_path)
    # Resize the input image to the network's input size
    # Normalize the image and turn it to Tensor
    transform1 = transforms.Compose([transforms.Resize(size=(224,224)),
                        transforms.ToTensor(),
                        transforms.Normalize((0.485, 0.456, 0.406), 
                                             (0.229, 0.224, 0.225))])
    # discard the transparent, alpha channel (that's the :3)
    img = transform1(img)[:3,:,:].unsqueeze(0)
    if use_cuda:
        img = img.cuda()
    # get model percentage outputs
    prediction = model_transfer.forward(img)
    # get highest prediction (index)
    prediction = prediction.cpu().detach().numpy() if use_cuda else prediction.detach().numpy()
    prediction = np.argmax(prediction, axis=-1)
    return class_names[prediction[0]]

Step 5: Write your Algorithm

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and human_detector functions developed above. You are required to use your CNN from Step 4 to predict dog breed.

Some sample output for our algorithm is provided below, but feel free to design your own user experience!

Sample Human Output

(IMPLEMENTATION) Write your Algorithm

In [147]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.
from matplotlib import rcParams
def run_app(img_path):
    ## handle cases for a human face, dog, and neither
    # using the previously made functions to detect if a face is detected or dog detected
    title = ""
    prediction = predict_breed_transfer(img_path)
    if dog_detector(img_path):
        title = "Hello doggo! you look like... %s breed." % prediction
    elif face_detector(img_path):
        title = "Hello human! you look like... %s breed." % prediction
    else:
        title = "I have no idea what I am looking at :)"
        
    prediction = prediction.replace(" ", "_")
    example = glob("/data/dog_images/train/*" + prediction)[0] + "/" +os.listdir(glob("/data/dog_images/train/*" + prediction)[0])[0]
    
    rcParams['figure.figsize'] = 20 ,20
    
    image = plt.imread(img_path)
    ex_image = plt.imread(example) 
    fig, ax = plt.subplots(1,2)
    ax[0].imshow(image)
    ax[0].axis('off')  # clear x-axis and y-axis
    ax[1].imshow(ex_image)
    ax[1].axis('off')  # clear x-axis and y-axis
    
    ax[0].set_title(title)
    ax[1].set_title("Here is a %s from our database" % prediction.replace("_", " "))
    plt.show()
    print("-" * 100)
    return

Step 6: Test Your Algorithm

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

Answer: (Three possible points for improvement) The output is better than expected, the testing accuracy might seem a little bit low (71%) but it definitely exceeds the average non-expert human prediction, it predicts from 133 breeds with many breeds looking alike. 1- I think to improve the model more the dataset should be increased since it's very small, using a bigger dataset would lead to a drastically higher test accuracy in my opinion. 2- I think using a deeper fully connected layer after having a bigger dataset would lead to higher accuracy, I tried using a deeper fully connected layer but the model started overfitting very quickly despite me using dropout and data augmentation. 3- Increase the number of epochs to train on after the previous two points are fulfilled.

In [150]:
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.
print("-" * 100)
print("My own samples images")
print("-" * 100)
print("-" * 100)
#### Test on humans not from the dataset ####
run_app('my_photo.jpg')
run_app('matthew.jpg')
run_app('dreymon.jpg')
#### Test on dogs from the internet ####
run_app('golden_retriever.jpg')
run_app('brussels_griffon.jpg')
run_app('boston_terrier.jpg')

### Test from random samples from our database ###
print("Samples from dataset images")
print("-" * 100)
print("-" * 100)
n = 2  # for 2 random indices
hindex = np.random.choice(human_files.shape[0], n, replace=False)
dindex = np.random.choice(dog_files.shape[0], n, replace=False)
human_files_sliced = []
dog_files_sliced = []

for i in range(2):
    human_files_sliced.append(human_files[hindex[i]])
    dog_files_sliced.append(dog_files[dindex[i]])
## suggested code, below
for file in np.hstack((human_files_sliced, dog_files_sliced)):
    run_app(file)
----------------------------------------------------------------------------------------------------
My own samples images
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
Samples from dataset images
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------------